<p>
</p>
<p>Yeah, but the point of a poll is to predict whether there is a real preference in the entire voting population, not whether someone gets the most votes among those polled. The point of statistical significance is whether the change you see between two groups could be due to chance. If it’s not statistically significant, you cannot say that the difference between two groups was not due to chance. In the case of a poll that isn’t statistical significant, you cannot conclude that if Romney leads Palin in a poll that Romney would actually win the primary.</p>
<p>On the other hand, the cut-offs for statistical significance are arbitrary, and it’s also possible that not enough data points were analyzed and the total N was too low. So there could be a real effect. I’d have to see the numbers to see how large the effect was and how close it was to being statistically significant.</p>
<p>I didn’t read the study in question, but another issue is how they treat the early action pool vs. the regular action pool. It sounds like they normalize to test scores and GPA, but it’s possible that the early action people spent more time on their essays (that’s a cliche’ I know, but in general I would think MIT early action people would care more about the MIT application.) Also, it takes some confidence to apply early action, so their EC’s might have been stronger. MIT applicants tend to saturate the upper range of SAT scores, so the early action and regular action pool could look like they are of equal academic strength even if they aren’t. If there is a small effect shown in the study, one of these factors could explain it.</p>